JP2020527303A - Bit plane coding - Google Patents

Abstract

The improvement in coding efficiency is that the coefficient groups for which a set of coded bit planes is predictively notified in the data stream are grouped into two group sets, and for each group set, all of each group set. Coding a Factor Group Performing bitplane coding in such a way that it is spent on a data stream that tells if the set of bitplanes is empty, that is, if all the coefficients in each groupset are not significant. Achieved by.
According to another aspect, the improvement in coding efficiency is achieved by providing bit plane coding with non-significant notification per groupset according to the first aspect as an alternative coding option to the notification of the groupset. , It is informed that there is no coded predictive residual of the coded bit planes of the claims groups in each groupset.

Description

The present application relates to bitplane coding such as bitplane image coding for still image and / or video coding.

Bitplane coding seeks to reduce the amount of coding required by limiting the coded bitplane to a portion of the total amount of available bitplanes.
In most cases, bit plane coding is performed on the transformation factor, i.e., the coefficient of transformation of the actual data to be encoded, such as the spectral decomposition transformation of the image.
Such a conversion already "condenses" the entire signal energy into a smaller sample, the conversion factor, and has the most significant bitplane of the available bitplanes, that is, the non-zero bits in each conversion factor. As far as the position of the most significant bit plane is concerned, it results in adjacent conversion coefficients that share similar statistics.
Therefore, in the currently envisioned version of PEG XS currently planned, the conversion factor representing the image is called GCLI, where the bits of the conversion factor within that group are occupied by the largest coded line index. The data stream is encoded in groups of conversion coefficients, spending syntax elements on each conversion coefficient group that represents the largest, i.e., highest-level bit plane.
The alternative name is the MSB position or the number of bit planes.
This GCLI value is encoded in the data stream in a predictive way, such as using spatial predictions from adjacent transformation factor groups.
Such GCLI groups are then grouped into SIG groups in order, and for each such SIG group in the GCLI group, the flag has a predicted residual encoded for the GCLI value within the SIG group. Spent in a data stream notifying if all zeros in all GCLI groups.
If such a flag indicates that all predicted residuals of GCLI within the SIG group are zero, then it is not necessary to send the predicted residuals of GCLI and the bit rate is preserved.
However, there is a continuing desire to improve the coding efficiency of just-lined bitplane concepts, for example in terms of compression and / or coding complexity.

An object of the present invention is to provide a more efficient bit plane coding concept.
This object is achieved by the subject matter of the independent claims.

According to the first aspect, the present application performs bit plane coding in such a way that the set of coded bit planes is predictively notified in the data stream and the coefficient groups are grouped into two group sets. If so, and for the group set, is the set of coded bit planes for all the coefficient groups in each group set empty, that is, is not all the coefficients in each group set significant? It is based on the finding that improved coding efficiency can be achieved when spent on data streams that inform.
This method encodes a set of coded bitplanes of a coefficient group in a particular groupset when nonetheless all the transformation coefficients in all the coefficient groups in a particular groupset are significant. It is possible to avoid spending unnecessary bits on the non-zero predicted residuals to do so, which tends to improve compression.
Beyond this, as far as the encoder is concerned, the determination of whether the conversion factors are not significant, i.e. the set of encoded bit planes, i.e. all non-zero bit planes, is below the quantization threshold. It can be determined in parallel with the set, i.e. independently of each other, which makes it easy to render the parallel implementation with non-significant notifications for each set of groups.

According to another aspect of the present application, the bit-plane encoding by non-significant notification per groupset according to the first aspect described above is for each groupset described in the introductory part of the present specification. Coding efficiency when provided as an alternative to the coding options associated with groupset notification, which can signal that there is no coded predictive residual in the coded bitplane of the claims group in the groupset. It turns out that the improvement can be achieved.
In contrast, according to the second aspect, the data stream consists of a first subset of the groupset in which the significant coding mode is not used and a second subset of the groupset in which the significant coding mode is used. Provides information to identify.
The first subset of the groupset is "successfully" encoded.
That is, the data stream provides the predicted residuals of the coded bit planes of the coefficient groups of such a group set, and where significant, the bits in the coded bit plane are encoded in the data stream.
In the second subset of the groupset, the data stream contains an indication or designation of a significant coding mode.
In other words, this instruction or designation informs the decoder of how to handle the second subset of the groupset, or, in other words, how to interpret the identification of the second subset of the groupset.
The first mode of significant coding mode corresponds to the interpretation that the predicted residuals of the coded bit plane notifications of the coefficient groups in such a groupset are zero.
For this purpose, according to the first significant coding mode type, only the predicted residual notification of the coded bit plane notification is omitted for the second subset of the groupset.
If the significant coding mode is indicated to be the second mode, the groupset of the second subset is treated as a collection of non-significant coefficients.
For this purpose, the decoder inherits that the coefficients are not significant for each coefficient group in such a groupset.
According to this second aspect, the encoder is provided with the opportunity to switch between both significant coding mode options, and the encoder can take advantage of this degree of freedom to select a coding mode that leads to higher coding efficiency.
However, beyond this, by providing the data stream with an opportunity to inform the decoder which significant coding mode option was used, the significant coding mode is more suitable for the encoder-side design and the encoder-side implementation of purpose. Provide an opportunity to choose an option.
For example, if there is a high interest in achieving higher parallelism, a non-significant signal mode, the second mode, may be preferred, but for single-threaded implementations of the encoder, the first mode may be preferred. May be done.
That is, the encoder can be implemented to operate only in either of both mode types selected to suit the encoder implementation.
Advantageously, the complexity of the decoder does not differ significantly between both mode types of significant coding modes.

An advantageous aspect of the present application is the subject matter of the dependent claims.
Preferred embodiments of the present application are described below with respect to the drawings.

It is a block diagram of the decoder operation of JPEG XS currently assumed. It is the schematic which shows the decomposition into the conversion coefficient of image 12 using the wavelet transform, that is, the conversion coefficient as an example of the object of bit plane coding. It is a schematic diagram showing the bitplane coding of the conversion coefficients in coefficient group units using the prediction-based coding of a set of coded bitplanes by GCLI delta coding. It is a schematic diagram which shows the subdivision of a wavelet transform into a pre-synch. It is a schematic diagram which shows the structure of the group set from the coefficient group in order to show the significance coding mode. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. Bitplane coding using RSF, ie coded bitplane notification, coding mode notifying the zero predicted residuals of the CSF, coding mode notifying the significance of all zeros of the associated coefficients, or the like. It is a figure which shows the PSNR simulation result obtained by the combination, that is, when both coding modes are switchable between encoders. It is a block diagram of the encoder by one Embodiment. It is a block diagram of the decoder by one Embodiment. It is a figure which shows the encoder by one Embodiment that the encoder uses the CSF coding mode. It is a block diagram of a decoder suitable for the encoder of FIG. FIG. 6 is a block diagram of an encoder using the RSSF coding mode that forms a data stream showing the use of the RSSF coding mode. FIG. 5 is a block diagram of a decoder capable of processing a data stream indicating either CSF or RSF use as an instruction. It is a figure which shows the pseudo code for explaining an example for forming a data stream which contains the instruction of use of RSF or CSF.

The following description of embodiments of the present application begins with a brief presentation of the current state of the JPEG XS standardization process, the version currently being discussed for JPEG XS, resulting in the embodiments of the present application. To be an overview of how this version can be modified.
Hereinafter, these embodiments are extended to provide additional embodiments described separately, but include individual references to the particular details described above.

FIG. 1 provides an example of a decoding process currently envisioned for JPEG XS to which embodiments of the present application can be applied, as described below.
As will become apparent from the extended embodiments described below, the application is not limited to this type of decoding process and the corresponding coding process.
Nevertheless, FIG. 1 assists one of ordinary skill in the art in gaining a better understanding of the concepts of this application.

According to FIG. 1, the code stream decoding includes the parsing part of block 1, the entropy decoding stage consisting of multiple blocks 2.1 to 2.4, the inverse quantization of block 3, the inverse wavelet transform of block 4, and so on. It is grouped into the inverse plural component uncorrelation of block 5.
At block 6, sample values are scaled, DC offsets are added, and they are clamped to the nominal range.

In block 1, the decoder parses the syntax of the code stream to get information about the layout of the sampling grid, as well as the so-called slice and presync dimensions.

Subpackets of the entropy-encoded data segment of the code stream are decoded into significant information, code information, and MSB location information (also called GCLI information), and all this information is used to become wavelet coefficient data.
This operation is performed in blocks 2.1 to 2.4 of FIG.

Image and video compression usually applies the transformation before performing entropy encoding.
For example, reference [7] uses block-based prediction, and references [4], [3], [5], and [6] propose wavelet transforms.
Wavelets are used in the case of FIG. 1, but FIG. 1 is just an example, and the same applies to the use of the wavelet transform.

Such a wavelet transform is shown in FIG.
Break the image into several subbands.
Each subband represents a spatially downsampled subband-specific spectral bandpass filter version of image 12.
As shown in FIG. 2, the number of horizontal decompositions may differ from the number of vertical decompositions.
At each decomposition step, the lowpass subbands of the previous decomposition are further decomposed.
For example, the L5 subband represents a subsampling version of the image, and the other subbands contain more information.

After frequency conversion, the subband coefficients are entropy encoded.
In other words, the g ≧ 1 coefficient of the subband ABm is A, B ∈ {L, H}, m ∈ N and is arranged in the coefficient group.
The top nonzero bit plane of the coefficient group is then notified, followed by the raw data bits.
Details of the coding technique will be described below.

FIG. 3 shows the principle of GCLI coding that ultimately becomes a syntactic element, the decoding of which corresponds to, for example, block 2.2 of FIG.
Therefore, GCLI coding relates to the coding of the most significant bit position and therefore to the indication of the coded bit plane.
This is done as follows. Several coefficients with more than one and having coefficients belonging to the same subband of frequency conversion are combined into groups, which are hereafter referred to as coefficient groups.
See, for example, FIG. The wavelet transform 10 drawn there is an example of the transform of the image 12.
Again, the wavelet transform is just one example of a transform to which the embodiments of the present application are applicable.
Instead of directly encoding the sample of image 12 or the value of pixel 14, the encoding is performed with a conversion factor of 16 in conversion 10.
FIG. 3 assumes that the coefficient group is composed of four coefficients.
However, the numbers are chosen solely for illustration purposes and may be chosen differently.
FIG. 2 shows, for example, that such coefficient groups 18 include four spatially adjacent transformation coefficients 16 all belonging to the same subband of transformation 10.
FIG. 2 shows that the coefficients 16 included in one coefficient group 18 are adjacent to each other in the horizontal direction, but this is also just an example, and the grouping of the coefficients 16 into the coefficient group 18 is performed differently. You may.
FIG. 3 shows a bit representation of each coefficient of the first coefficient group of reference numeral 20 and the second coefficient group of reference numeral 22 on the left side.
The bit of the absolute value of each coefficient extends along the column for each coefficient.
Therefore, the four columns are indicated by reference numerals 20 and 22, respectively.
Each bit belongs to a specific bit plane, the least significant bit in FIG. 3 belongs to the least significant bit plane, and the most significant bit belongs to the most significant bit plane.
For illustration purposes, eight available bitplanes are shown in FIG. 3, but the numbers may vary.
In addition to the magnitude bits 24 of the conversion coefficients, FIG. 3 shows the sign bit 26 above the corresponding magnitude bits for each coefficient.
GCLI coding will be described in detail with reference to FIG.

As already outlined, the coefficients are expressed in sine magnitude representation.
The maximum coefficient for each coefficient group determines the number of active bitplanes for this coefficient group.
A bitplane is called active if at least one coefficient bit 24 of the bitplane itself or more bitplanes (bitplanes representing a larger number) is non-zero.
The number of active bitplanes is given by the so-called GCLI value, the maximum coded line index.
For example, in the coefficient group 20, the GCLI is 6, and in the second coefficient group 22, the GCLI is 7, for example.
If the GCLI value is 0, then the complete coefficient group is 0 because there is no active bitplane.
This situation is known as non-significant GCLI, and significant GCLI is the opposite.
To achieve compression, only the active bitplane is placed in the bitstream, i.e. encoded.

Lossful coding may require truncation of parts of the bitplane so that the number of bitplanes transmitted for the coefficient group is less than the GCLI value.
This truncation is specified by the so-called GTLI, the largest trimmed line index.
The alternative name is the truncated position. Zero GTLI corresponds without truncation.
A GTLI value of 1 means that the number of transmitted bit planes in the coefficient group is one less than the GCLI value.
In other words, GTLI defines the smallest bitplane position contained in the bitstream.
For a simple dead zone quantization scheme, the transmitted bit plane is equal to the bit plane of the coefficient group without the truncated bit plane.
For more advanced quantization schemes, the quantization bin can be modified to "push" some information of the planned truncated bits into the transmitted bit plane.
Details can be found in [6].

For each coefficient, the number of residual bit planes is equal to the difference between the GCLI and GTLI values, so it is clear that the bitstream does not contain a coefficient group whose GCLI is less than or equal to the GTLI value.
In other words, the (data) bits 24 are not transmitted in the bitstream of these coefficient groups. Those coefficients are not significant.

The active bitplane that remains after truncation and quantization is referred to below as the residual bitplane, or truncated GCLI.
In addition, GTLI is also referred to below as the truncation point. If the residual bit plane is zero, GCLI is known as insignificant truncated GCLI.

These residual bit planes are transmitted to the decoder as raw bits.
Block 2.3 in FIG. 1 is responsible for deriving these bits from the bitstream.
However, to enable correct decoding, the decoder needs to know the GCLI values for all coefficient groups 18.
With the GTLI value also notified to the decoder, the decoder can estimate the number of raw data bitplanes in the bitstream.

The GCLI value itself is signaled by a variable length code that represents the difference from the GCLI value of the previous coefficient group.
This earlier coefficient group is, in principle, any coefficient group that the encoder has already previously decoded.
Thus, for example, it can be a horizontal or vertical adjacent group.
The output from the prediction is the difference in the number of residual bitplanes between the two coefficient groups, resulting in a delta residual bitplane.
For example, in FIG. 3, it is assumed that the coefficient group on the left side indicated by reference numeral 20 precedes in the order of coding of the coefficient group 22, and that the GCLI functions as a predictor of the GCLI of the coefficient group 22.
Details will be described below.
Note that GCLI values below the GTLI value are not of interest as their coefficients are not included in the bitstream in any case.
As a result, the prediction is performed so that the decoder can infer whether the GCLI is greater than the GTLI and, if so, the value of the GCLI.

Note that the method described below does not depend on the transmission order of the different bitstream parts.
For example, the GCLI coefficients of all subbands can be placed in the bitstream first, and then the data bits of all subbands can be placed.
Alternatively, the GCLI and the data bits may be interleaved into the data stream.

The frequency conversion coefficients shown in FIG. 2 are configured in so-called precincts 30. This is shown in FIG.
The presync groups groups the coefficients of different subbands that contribute to the predetermined spatial region 32 of the input image 12.

To be able to recover the signal, the decoder needs to know the GCLI values for all coefficient groups 18.
According to [3], there are various methods for efficiently notifying them.

In RAW mode, the GCLI value is transmitted unpredictably.

Therefore, let F _{1 be} the coefficient group to be encoded next.
The GCLI value can then be encoded with a fixed length codeword representing the following values:
max ^{f ()} (GCLI (F ₁ ) -GTLI (F ₁ ), 0)

In horizontal prediction, the encoded symbol is the difference between the GCLI value and the previously encoded GCLI value that belongs to the same line and the same wavelet subband and takes GTLI into account.
This difference value is hereinafter referred to as the residual or δ value.

Let F ₁ and F _{2 be} two horizontally adjacent coefficient groups composed of g> 1 coefficient. Let F ₂ be the currently encoded coefficient group. The GCLI (F ₂ ) can then be notified to the decoder by transmitting the residuals calculated as follows:

The decoder recovers GCLI (F ₂ ) by calculating the following equation.

Note that in horizontal predictions, normal GTLI (F ₁ ) = GTLI (F ₂ ). Furthermore, it should be noted that δ is transmitted as a variable length code as described in [4].

For vertical predictions between two subband lines, the result is the difference between the GCLI value and the GCLI of the same subset of the coefficients of the previously encoded line.

Let F ₁ and F _{2 be} two vertically adjacent coefficient groups composed of g> 1 coefficient. Let F ₂ be the currently encoded coefficient group.
The GCLI (F ₂ ) can then be encoded in the same way as the horizontal prediction.

Vertical prediction is limited within slices, where slices are a set of predetermined contiguous rows (eg 64 rows).
With this method, the first presync of the slice cannot be predicted vertically.

An alternative to vertical prediction is to use the following prediction formula instead of the above prediction.
δ = max (GCLI (F ₂ ), GTLI (F ₂ ))-max (GCLI (F ₁ ), GTLI (F ₁ ))

ここで、

であり、
ｇ_ｉは符号化するＧＣＬＩであり、

は基準として使用されるＧＣＬＩであり、
ｔ_ｉはｇ_ｉに適用する切り捨てであり、

は

に適用された切り捨てである。 Another alternative to vertical prediction is to use so-called bounded codes.

here,

And
g _i is a GCLI to be encoded,

Is the GCLI used as a reference,
t _i is the truncation to be applied to the _{g i,}

Is

Truncation applied to.

Since such a code has a characteristic of δ ≧ 0, efficient simple coding is possible.

The same prediction method can be applied to the following equation.

In [1], an escape code is used in GCLI coding to notify a sequence of coefficient groups, all of which are composed of a plurality of coefficients smaller than a predetermined truncation threshold.
By these means, a plurality of zero coefficient groups can be represented by one escape word without requiring a code word for each coefficient group, so that the coding efficiency can be improved.

This method has the advantage of not requiring any overhead for the significance flag, but it is a bit more complicated to calculate the additional bits compared to the bits required without the escape code.
Moreover, some encoding methods do not allow the use of escape codes in a simple way.

For example, FIG. 5 shows a coefficient group 18 that can be immediately contiguous with respect to the coding sequence 38 defined within the coefficient group 18, but this is not required.
According to the escape coding above, the GCLI values transmitted in the data stream of coefficient group 18 assume an escape code and together with that coefficient some subsequent coefficient groups 18 that form a groupset 40. It can be notified that all the coefficients of are not significant.
Questions about which coefficient group 18 belongs to groupset 40 may be known or notified by default.
For example, if the GCLI values of consecutive GCLI coefficient groups 18 are slightly truncated, they can be discarded from the code stream to improve coding efficiency, for example, in this way.
In this spatial zero-run method, this is done by encoding the escape value of the first coefficient of the insignificant groupset 40.
However, as shown earlier, the use of such coded escape values increases the complexity of the coding and is not suitable for extremely low complexity.

According to the so-called RSF method taught in [1], the burden of coding the GCLI value is that the non-significant truncated GCLI value of the coefficient group 18 is 0 from all the reference GCLI values in the groupset 40. It is alleviated by notifying a groupset such as the groupset 40 of FIG. 5 that it is predicted to reach a residual equal to.
For this purpose, the coefficient groups 18 are grouped into groupset 40 and the data stream contains, for each groupset 40, an RSF flag indicating whether the predicted residuals of GCLI are all 0 within the groupset 40. , In that case, of course, there is no predictive residual that needs to be transmitted in the data stream.
However, RSF does not skip non-significant GCLI coding if the corresponding residual is non-zero.

The predicted residuals of GCLI for set 40 may not be zero, but due to truncation, all coefficients in each coefficient group 18 within each group set 40 are not significant.

The embodiments described below provide an opportunity to remove insignificant truncated GCLI from the code stream by modifying the interpretation of RSF, making it possible to complement the RSF method outlined in low complexity. To do.

This will be described in more detail below.

In the RSF method described in [1], the GCLI coefficients are arranged in groups within each subband, and are hereinafter referred to as SIG groups.
For example, element 40 in FIG. 5 is such a SIG group. The size of the SIG group is 8 or any other number greater than 1.
That is, the SIG group 40 may include two or more coefficient groups 18.
Coefficient group 18 included in one SIG group 40 may belong to one subband of transformation 10 as outlined, but this is not required.
Note that the last factor may be treated as an incomplete group if the subband is not a multiple of the SIG group size, such as 8.

For example, at the start of the presync 30 code stream, a series of flags is notified.
Each flag corresponds to each pre-synced SIG group 40.
When the flag is set, it means that it does not exist in the code stream because all the GCLI residuals corresponding to the group 40 are 0.

As mentioned above, there are situations where the GCLI of the SIG group is completely truncated (or simply goes to zero), while the residual is not zero.
This can occur, for example, when GCLI is predicted vertically from a significant line or row.
Here, for example, if a non-zero residual requires more budget for binary coding, it may actually be advantageous, but RSF does not succeed in preventing notification of the residual.

Therefore, instead of RSF, the Coefficient Significance Flag (CSF) is used according to embodiments of the present application, thereby aiming to further extend the definition of RSF.
By introducing a new GCLI coding method, the CSF also assigns one flag to all SIG groups 40, but is always set if the GCLIs of coefficient group 18 of SIG group 40 are not all significant after truncation, ie. , The set of coded bit planes for these coefficient groups 18 is empty.
Therefore, the same amount of flags as in RSF is needed.
As described below, CSF coding can be combined with RSF coding in the sense that both can be used according to alternative coding options, for example presync 30 or per subband.
Here, the same flags in the data stream are interpreted as RSF or CSF depending on the additional signaling in the data stream.

Table 1 shows examples of four typical SIG groups and how RSF and CSF are set up in these SIG groups.

Table 1 shows examples and comparisons of the CSF and RSF methods.
In the case of SIG group 0, CSF is selected because all truncated GCLI values are 0, but the RSF flag does not indicate that the residual is not 0.
In the case of SIG group 1, the situation is the opposite.
In the case of SIG group 2, since both GCLI and the residual are 0, the CSF becomes 1 and the RSF also becomes 1.
Finally, in SIG Group 3, neither is selected, i.e. RSF and CSF are set to zero.

The CSF variant will be further described below.

For example, using the CSF flag has the effect of saving budget for each SIG group.

The simple-ary coded alphabet usually signals a residual value of 0, leaving 1 bit behind.
Therefore, the budget saved by RSF is always the same for all deleted SIG groups, exactly the size of the group.
On the other hand, the budget overhead introduced by this method is constant across the image and is always equal to the amount of RSF required.

For CSF, the budget overhead is exactly the same as for RSF.
However, in contrast, peak budget savings per SIG group are equal to or better than RSF. In fact, the residuals removed by the CSF can be equal to or different from zero, so the budget can be larger than the size of the group.

Although RSF can be used transparently for prediction, the decoder and encoder prediction modules are slightly modified because CSF is post-processing (in the encoder) or pre-processing (in the decoder).

If the encoder finds that the SIG group contains only non-significant truncated GCLI, its coding can be skipped altogether.
However, to get the amount of bits saved by the residuals, you need to calculate further in the budget calculation.
The decoder can also skip the inverse prediction of GCLI deleted by CSF and replace it with 0 instead.

The outline of image coding using CSF will be described in detail below.
For this reason, some function definitions are used as follows.

Let α be a coefficient group that encodes.
GCLI (α): Returns the GCLI value coefficient group a PREF (α): Returns the coefficient group used to predict the GCLI value of the coefficient group α GTLI (α): Returns the GTLI value applied to the coefficient group α.
GTLI (α) depends on the pre-synced and sub-bands of group α.
PRED (g _α , t _α , g _b , t _b ): Using GCLI g _b as a reference, returns the residuals corresponding to the prediction of GCLI g _α . GTLI is t _α and t _b , respectively.
PRED ^-1 (δ, t _α , g _b , t _b ): Returns the inverse prediction corresponding to the coefficient group using GCLI g _b as a reference and residual δ.
GTLI is t _α and t _b , respectively.
SIZE (s): The number of coefficient groups in one line of subbands s.
SIGGRP (α): Returns the index of the SIG group to which the coefficient group α belongs.
CSF (g): Returns whether the CSF flag is associated with the SIG group g. True means that the group is not significant.
RSF (g): Returns whether the RSF flag is associated with the SIG group g. True means that the group is not significant.
SIGSIZE (s): The size of the SIG group, which may be different for each subband.

The pseudo code for managing CSF is shown below.

Decoding of the GCLI value of the subband is performed as follows.

When using CSF, the decoder can be described as:
For the subband s, the set GCLI ( _ai ) of values for the coefficient group _ai is decoded as follows:
for 0 ≤ i <SIZE (s)
if CSF (SIGGRP ( _ai )) # If the STG flag is set, there is no GCLI GCLI (ai) ← 0
If not the else # likely decodes the GCLI δ '= vlc_decode () # GCLI to unary decoding the delta values ^{_{(a i) ← PRED -1 (}} δ', GTLI (a i), GCLI (PREF (a i )), GTLI (PREF ( _ai )))
end if
end for

The coding of the GCLI value of the subband is performed as follows.

We define a GTLI in which all GCLIs in the SIG group are not significant as follows.

That is, it is the maximum GCLI value of the group.
Therefore, the coding of the coefficient group _ai of the subband s can be performed as follows.
for 0 ≤ i <SIZE (s)
if GTLI ( _ai ) ≥ GTLI _csf (SIGGRP ( _ai )) # In case of insignificant truncated GCLI CSF (SIGGRP ( _ai )) ← True # Just update the flag, do not encode else # Not In the case of encoding (PRED (GCLI ( _ai ), GTLI ( _ai ), GCLI (PREF ( _ai )), GTLI (PREF ( _ai ))))
CSF (SIGGRP ( _ai )) ← False
end if
end for

In comparison, the pseudo code for managing RSF is shown below for reference.

First, the decoding of the subband GCLI value is checked.

When using RSF, the decoder can be described as:
For the subbands s, the set of values for the coefficient group ( _ai ) GCLI ( _ai ) is decoded as follows:
for0 ≤ i <SIZE (s)
if RSF (SIGGRP ( _ai )) # If the STG flag is set, there is no GCLI δ'← 0
else # If not, decode GCLI δ'= vlc_decode () # Digit decode delta value end if
GCLI ( _ai ) ← PRED ^-1 (δ', GTLI ( _ai ), GCLI (PREF ( _ai )), GTLI (PREF ( _ai )))
end for

The coding of the subband GCLI value when RSF is used is as follows.

We define a GTLI in which all residuals in the SIG group are not significant as follows.

Therefore, the coding of the coefficient group _aj of the subband s can be performed as follows.
for 0 ≤ i <SIZE (s)
if GTLI ( _ai ) ≥ GTLI _rsf (SIGGRP ( _ai )) # In case of insignificant truncated GCLI RSF (SIGGRP ( _ai )) ← True # Just update the flag but do not encode Else # Not In the case of encoding (PRED (GCLI ( _ai ), GTLI ( _ai ), GCLI (PREF ( _ai )), GTLI (PREF ( _ai ))))
RSF (SIGGRP ( _ai )) ← False
end if
end for

It is possible to support switching between the coefficient flag and the residual significant flag.
As described above, the coefficient significance flag can indicate the existence of a series of coefficient groups (so-called SIG groups) that become zero after quantization even if the predicted residuals are not zero.
Placing a codeword representing a predicted residual in a bitstream can be avoided by correspondingly setting significant information or a significant flag representing a sig group, thus improving coding efficiency.

On the other hand, the residual significance flag indicates the existence of a quantized SIG group with all zero predicted residuals.
In other words, if all the quantization coefficients of the SIG group have the same predicted values that may be different from zero, then the zero predicted residuals if the corresponding significant bits of the SIG group are set appropriately. There is no need to place the difference in the bitstream.
Therefore, the bitstream of all presyncs (or all subbands) indicates which of the two significant flags is selected.
These means allow the encoder to select the best alternative for all presyncs or all subbands and provide coding gain as described below.

Figures 6.1 to 6.32 are GCLI codes using the above coding framework, ie in combination with variants that allow switching between RSF coding, CSF coding, or both coding modes. The PSNR results obtained using the chemicals are shown.
Figures 6.1 to 6.6 are optimized for RGB 4448-bit coding, ie RSF / CSF switchability, with various bpp (bits per pixel) constraints, RSF only, CSF. Only, PSNR only, RSF only, CSF only, RSF / CSF switching visually optimized with 4 bpp bit rate constraint while comparing RSF / CSF switching respectively, optimal with 6 bpp bit rate constraint Optimized with a bit rate constraint of 12 bpp, comparing PSNR, CSF only, CSF only, and RSF / CSF switching, respectively, and 4 bpp bits while comparing RSF only, CSF only, RSF / CSF switching, respectively. The PSNR is visually optimized with a rate constraint and shows a visually optimized PSNR with a bit rate constraint of 12 bpp, comparing RSF only, CSF only, and RSF / CSF switching respectively.
Similar simulation results are shown in FIGS. 6.7 to 6.10 for -RGB 444 10-bit coding and in Figures 6.11 to 6.14 for YUV 422 10-bit coding. It is shown in.
The results of the maximum multiple generation PSNR and the average PSNR are shown in FIGS. 6.15 to 6.32.
That is, RGB 444 8 bits of FIGS. 6.15 to 6.20, RGB 444 10 bits of FIGS. 6.21 to 6.26, and YUV 422 10 bits of FIGS. 6.27 to 6.32.

In the following, some complex aspects related to CSF and RSF will be described.
However, previously, the encoder architecture has been presented with reference to FIG.
The encoder is shown in FIG. 7 using reference numeral 50 in a manner that uses the wavelet transform 10 described above as a starting point.
The wavelet transform 10 may be obtained by a wavelet transform by a converter not shown in FIG.
To encode the wavelet transform 10, the encoder 50 includes a GCLI extractor 52 that determines the maximum encoded line index for each coefficient group 18.
The encoder 50 operates, for example, on a pre-synced basis and operates to meet a specific constrained bit rate.
The GCLI extractor 52 supplies the determined GCLI value to the GCLI buffer 54 and the execution / GTLI module 56.
Module 56 calculates the minimum GTLI and truncates all coefficient groups of the significant groups to zero.
Details will be described below.
Module 56 transfers the GCLI value and the smallest GTLI leading to an insignificant significant group to a subsequent budget computer 58 that calculates a bit budget for each GTLI candidate value.
For this purpose, module 58 continues to access and update the conversion coefficient line that acts next in the coding order on the GCLI value of the previous conversion coefficient line in buffer 60.
Module 58 only calculates an initial approximation of the bit budget for each GTLI candidate, as the exact budget depends on the previous pre-synced GTLI that may not yet be available. More precisely, the budget computer 58 operates on the current pre-synced conversion factor line.
The pre-synced budget updater 62 connected to the GCLI buffer 54 provides a pre-synced budget update.
Correct the deviation between the previous budget approximation and the actual bit budget.
For this purpose, module 62 operates on the presync and is then encoded by module 64 so that the previous presync GTLI is already available.
Based on the pre-synced budget update and the budget value determined by the computer 58, the RA module 63 calculates the GTLI value and applies it effectively to the pre-synced, then encodes to satisfy the bitrate constraints described above.
This GTLI value is provided as an input to the GCLI coder 64, and further receives the GCLI value from the GCLI buffer 54.
The GCLI coder 64 accesses the GTLI of the previous line and the previous line of the GCLI value in the form of the previous line GCLI buffer 60 via the register 66.
The GCLI coder 64 encodes the GCLI value with the details set above in this regard and outputs it to buffer 68.
The coefficients of conversion 10 are also buffered in buffer 70, and those bits in the encoded bit plane notified in the data stream via the encoded GCLI value in buffer 68 are inserted into the data stream via the coefficient encoder 72. ..
As mentioned above, this can be done by placing the bits as raw data in the data stream.
The packer 74 packs the encoded GCLI data and raw data bits into a data stream.

The highlighted blocks of FIG. 7, reference numerals 56, 58, and 64, relate to the use of two different types of significance flags: RSF and CSF.
The wavelet coefficient of the wavelet transform 10 is stored in the coefficient buffer 70 for later data coding by the encoder 72, as further described in [1], the contents of which are incorporated herein by reference.
As described in [1], the GCLI extractor 52 determines the GCLI value and stores it in the buffer 54.

To combine two different sigflag methods, it is necessary to calculate the following values for each significant group, such as by module 56.

は
ｇ_ｊ
を予測するために使用される基準ＧＣＬＩ（すなわち、水平または垂直の近傍）である。
・

はＧＣＬＩ

に適用する量子化／切り捨てである。
・ＰＲＥＤは、ｔと

を使用して

から値ｇ_ｊを予測する予測関数である。
・ｓ_ｓｉｇは現在処理されている有意グループである。 here,
-G _j is GCLI j.
• t _j is the quantization / truncation applied to GCLI g _j .
・

Is g _j
Is the reference GCLI (ie, horizontal or vertical neighborhood) used to predict.
・

Is GCLI

Quantization / truncation applied to.
・ PRED is t

using

It is a prediction function that predicts the value g _j from.
• s _sig is a significant group currently being processed.

は［２］で使用される値と同じ値であるため、複雑さについてはこれ以上説明しない。

の計算は、コンパレータ（＜＝５ＬＵＴ）およびサブバンドごとに４ビットの１つのレジスタによって可能である。
さらに、ＧＣＬＩ値を１つの有意グループだけ遅延させることにより、初期予算計算が簡素化される。
１つの垂直ウェーブレット分解レベル（３×８サブバンド）の場合、これには３×８×８×４＝７６８ビットが必要である。
ザイリンクスの場合、これは２×４８＝９６ＬＵＴ、またはアルテラデバイスの２ ＭＬＡＢブロックに対応する。

Is the same value used in [2], so complexity will not be discussed further.

The calculation of is possible with a comparator (<= 5 LUT) and one register with 4 bits per subband.
In addition, delaying the GCLI value by one significant group simplifies initial budgeting.
For one vertical wavelet decomposition level (3x8 subband), this requires 3x8x8x4 = 768 bits.
For Xilinx, this corresponds to 2 x 48 = 96 LUT, or 2 MLAB blocks for Altera devices.

を使用する場合、コーダは、符号化するすべてのＧＣＬＩ ｇ_ｉが選択した量子化／切り捨てパラメータｔ_ｉをすべて下回っているかどうかをさらに確認する必要がある。
しかし、これは簡単であり、追加のバッファリングは必要ない。 The GCLI coder needs another minor change.
When using only the residual significance flag (such as [1]), it is necessary to buffer the s _sig predicted residual before encoding to determine if all are zero.
This allows you to output the predicted residuals or use the significance flag to notify the SIG group that it is not significant.

When using, the coder needs to further determine whether all GCLI g _i to be coded is below all of the quantization / truncation parameter t _i selected.
However, this is easy and does not require any additional buffering.

The calculation of the budget saving of the coefficient significance flag is performed as follows.

Each time the coefficient significance flag is used to signal that the significance group is not significant, the budgeting module of FIG. 7 needs to track the number of bits saved by not placing the predicted residuals in the bitstream.

Therefore, the overall budget for both methods can be calculated as:
-Calculate budget without significance flag-Calculate budget savings for residual significance flag-Calculate budget savings for coefficient significance flag

This means that the added complexity of using both methods, as explained below, lies in the calculation of additional budget savings.

It is assumed that the vertical prediction by the first option described above is applied.

（１）
現在のＧＴＬＩと基準ＧＴＬＩのｔ_ｉおよびｔ_ｉ−１の両方が等しい場合、式（１）は次のように単純化される。

（２）

予算の節約は

についてのみ発生することがわかっているため、式（１）から次式を得る。
δ_ｉ＝ｔ_ｉ−ｍａｘ（ｇ_ｉ−１，ｍａｘ（ｔ_ｉ−１，ｔ_ｉ））
したがって、予算の節約は、ｇ_ｉ−１と

、さらにパラメータｔ_ｉ−１とｔ_ｉに一意に依存するため、簡単に計算できる。 The following equation is used in this prediction method.

(1)
If both _{t i} and _{t i-1} of the current GTLI and reference GTLI are equal, equation (1) is simplified as follows.

(2)

Budget savings

Since it is known that this occurs only in the above equation (1), the following equation is obtained.
_{δ i = t i -max (g} i-1, max (t i-1, t i))
Therefore, the budget savings are g _i-1

, To uniquely dependent on further parameters _{t i-1} and _{t i,} can be easily calculated.

現在のＧＴＬＩと基準ＧＴＬＩのｔ_ｉおよびｔ_ｉ−１の両方が等しい場合、式（１）は次のように単純化される。

（４）
予算の節約は

についてのみ発生することがわかっているため、式（３）から次式を得る。

したがって、予算の節約は、ｇ_ｉ−１と

、さらにパラメータｔ_ｉ−１とｔ_ｉに一意に依存するため、簡単に計算できる。 If the second vertical prediction option applies, then the following equation is used:

If both _{t i} and _{t i-1} of the current GTLI and reference GTLI are equal, equation (1) is simplified as follows.

(4)
Budget savings

Since it is known that this occurs only in the above equation (3), the following equation is obtained.

Therefore, the budget savings are g _i-1

, To uniquely dependent on further parameters _{t i-1} and _{t i,} can be easily calculated.

The corresponding decoding architecture is shown in FIG.
The decoder of FIG. 8 is generally shown using reference numeral 80.
The input demultiplexer 82 receives the data stream and derives from it the coding coefficient bits in the coding bit plane, ie code 84, the GCLI residual 86, and the flag, which may be RSF or CSF, ie code 88. To do.
Bit 84 is stored in the data buffer 90, the GCLI residual value is stored in the GCLI buffer 92, and the flag 88 is stored in the buffer 94.
As shown in FIG. 8, the GCLI residual value stored in the buffer 92 is encoded as monochromatic or raw data, and the raw decoder 96 or the monochromatic decoder 98 decodes the GCLI residual value accordingly. Used to do.
The subband GCLI buffer 100 may optionally have access to the decoder 96 and the decoder 98, respectively.
Through the GCLI packer 102, the inverse GCLI predictor 104 receives the GCLI residuals and reconstructs the GCLI values based on the previous GCLI 106 and based on the RSF flag.
When RSF is applied and the current GCLI RSF flag is set, the inverse predictor 104 is notified of the predicted residual, i.e., the GCLI residual which is zero anyway.
The predictor determined based on the previous GCLI 106 is then used as the current GCLI.
The inverse predictor 104 outputs the determined GCLI, and the multiplexer selects this predictive output or zero substitution, depending on the CSF flag applied to the current GCLI.
If the CSF flag is set, there is no coded bitplane in the corresponding SIG group anyway, and the GCLI is set accordingly, i.e. the code of the conversion factor that is not significant given the current GTLI. It becomes zero or some value that leads to conversion.
The unpacker controller 110 receives the output of the multiplexer 108, which output is then fed back to the inverse predictor 104 as the previous GCLI, this time from the data buffer 90 to the current coefficient group, depending on the current GCLI. Controls the unpacker 112 that extracts the coefficient bits of the coded bit plane of.
At the output of the unpacker 112, each conversion factor is generated.

Therefore, FIG. 8 shows a block diagram of the decoder, especially in addition to [1], showing an extension that allows support for both significant flag types.
Note that for completeness, FIG. 8 shows an additional GCLI packer 102 compared to [1].

If the presync (or subband) is encoded with the residual significance flag, the inverse predictor simply considers the predicted residual to be zero rather than reading the predicted residual from the GCLI packer 102.
When using the coefficient significance flag, the inverse predictor 104 can perform exactly the same operation.
However, instead of using the result of this prediction, the value is simply replaced with a zero value.
Therefore, to handle both flag types, the decoder of FIG. 8 simply includes a 4-bit MUX2 element controlled by the output of the significance flag buffer 94 and the type of significance flag used, namely reference numeral 108.
Therefore, as far as the decoder is concerned, the increase in logic is negligible.

After describing certain embodiments of the present application as extensions or modifications to currently envisioned versions of JPEG XS, further embodiments of decoders and encoders and data streams will be described as a generalization of the above embodiments. ..
FIG. 9 shows the encoder 100. The encoder of FIG. 9 is for encoding the conversion coefficient 16 into the data stream 102.
As described above, the conversion coefficient 16 may be a conversion coefficient for image conversion.
The conversion factor 16 forms one sub-part of the plurality of sub-parts of the spectral decomposition of the image, and the encoder 100 may be configured to perform encoding on a sub-portion basis. Good.
Such a subband may be a subband, such as a wavelet transform subband, or a group of transform coefficients associated with the corresponding spatial region of the spatial region in which the image is subdivided, such as the region 32 corresponding to the presync. Good.
It is clear that the conversion factor 16 may be a different conversion factor, such as DCT.
The conversion coefficients are grouped into coefficient groups 18. As described above, the number of coefficients 16 per group 18 may be any number greater than 1 and is not limited to 4 shown in FIG.
Grouping of the coefficients 16 into groups 18 may be performed so that the coefficients 16 belonging to one group 18 belong to the same subband.
In the case of the wavelet transform, the coefficients 16 belonging to one group 18 may be, for example, spatially adjacent to one subband, and when the transform coefficient 16 is a DCT coefficient, the group 18 is spatially adjacent to the image. It may be composed of coefficients 16 derived from different DCT transform blocks obtained from the region, and the coefficients of one group correspond to one frequency component or coefficient in these DCT transform blocks.
In particular, in the case of DCT transform, the image can be transformed into DCT transform blocks of the same size in block units, and each coefficient position represents a separate subband.
For example, all DC coefficients of these DCT transform blocks represent a DC subband, a coefficient to the right of it, another subband, and so on.
Group 18 may then collect the coefficients of one subband of the DCT transform blocks obtained from adjacent blocks of the image.

The coefficient group 16 is then grouped into the group set 40.
This can also be done in a way that does not mix the coefficients of different subbands.
Further, the coefficients 16 of the coefficient groups 18 in one group set 40 may all originate from the same subband.

The encoder 100 of FIG. 9 is a first subset of the groupset 40 in which the significant coding mode is not used, i.e., a groupset 40 in which the GCLI residuals are encoded, and a second groupset in which the significant coding mode is used. Information 104 that identifies a subset of, i.e., the groupset 40 in which the GCLI residuals are unencoded, is inserted into the data stream 102.
In the above description, one CSF flag is inserted into the data stream 102 for each groupset 40 to form the information 104.
The first subset of the groupset 40 is the groupset 40 with a CSF of 0 or not set, and the second subset includes the groupset 40 with a CSF of 1.
To set the information 104 of the current groupset 40, the encoder 100 checks whether all the conversion coefficients 16 in the groupset 40 are insignificant, i.e. quantized to zero 106.
The encoder 100 can insert truncation information 108 indicating a set of one or more truncated least significant bit planes into the data stream 102.
The GTLI value described above can form part of information 108.
The GTLI 108 may be transmitted in the data stream 102, for example, at the particle size of the sub-parts described above, i.e., for example per presync, or at another level, such as in units of subbands or coefficient group rows.
As an aspect, it should be noted that the coefficient groups 18 may collect coefficients 16 adjacent to each other along a diagonal direction with respect to the coefficient row 41, in addition to being exemplified by the figures. The GCLI value for which information 118 provides the predicted residuals can indicate the most significant bit plane encoded in the data stream 102 as an index associated with GTLI, where GTLI is then truncated to the magnitude bit 24. The most significant bit in the least significant bit plane can be indexed.
If all the coefficients of all the groups 18 in the current groupset 40 are 0, then the CSF flag of this groupset 40 is set with code 110, otherwise it is set as shown by code 112. Not done.
If not set, the encoder 100 of the encoded bitplane by predicting this set at reference numeral 114 based on the adjacent coefficient group 18, eg, by inserting the predicted residuals into the data stream 102. The set is notified by the data stream 102, thereby forming the GCLI data 118 in the data stream 102.
For example, a set of coded bit planes may be signaled in the data stream 102 by indexing, i.e. by indexing the largest coded line.
For the coefficient groups 18 in the current group set 40 where the GCLI 118 is larger than the GTLI, the encoder 100 encodes the corresponding coefficient bits, or bits 24, of the coefficients 16 of the respective coefficient groups 18 into the data stream 102.
More specifically, the bit insertion 120 may be performed at a code rate of 1 by inserting a bit as a raw bit or the like.
Next, the GCLI data value may be encoded in the data stream 102 as a variable length code such as the above-mentioned simple code.
The raw bit inserted at reference numeral 120 is shown by reference numeral 122 in FIG.
As already mentioned above, in the data stream 102, the raw bits 122, the GCLI data 118, and the flag 104 may be interleaving or non-interleaving.
As shown, CSF = 1 is a highly compressed method representing groupset 40.
After either reference numeral 110 or reference numeral 120, processing can proceed in another groupset 40 in the same manner.

FIG. 10 shows a decoder corresponding to the encoder of FIG.
The decoder 200 of FIG. 10 operates to reconstruct the conversion factor 16 from the data stream 102 and checks if the CSF of the current groupset 40 is set for this purpose, in which case the decoder. The 200 sets all conversion factors 16 in the groupset 40 to zero, that is, sets them to zero, or synthesizes noise into the conversion factors 16.
For this purpose, if check 206 indicates that the significant coding mode is used for the current groupset 40, some non-significant processing 210 is performed for the current groupset 40. ..
However, if this is not the case, the decoder 200 processes the groupset 40 as usual.
That is, the decoder 200 predicts the GCLI of each coefficient group 18 in the current groupset 40 at 214 and modifies the prediction using the prediction residuals obtained from the data stream 102.
As mentioned above, variable length decoding may be used to derive the predicted residual 118.
The prediction can be made using the current coefficient group or the GCLI of the coefficient group 18 perpendicular to the current group 40.
That is, in this example, for all groups 18 in the current set 40, the prediction criteria, i.e., the vertically adjacent groups 18, are outside the current set 40.
Alternatively, prediction 214 may be made using the current coefficient group or the GCLI of coefficient group 18 horizontally adjacent to the current group 40.
That is, in the current example, for most of the groups 18 in the current set 40 except the leftmost, for example, the predictive criteria, that is, the horizontally adjacent groups 18, are in the current set 40.
Of course, it is feasible to notify the prediction source of each group 18 in the data stream.
It may be a mode that can even be unpredictable.
Of course, the details of prediction 214 can also be transferred to prediction 114.
The mode switch may be optionally notified and selected by the encoder at a particle size other than group 18 or set 40, such as coefficient row 41 or group 18 row, subband or predincts 30.

For each coefficient group 18 whose GCLI is greater than GTLI, i.e. the set of coded bit planes is not below the quantization threshold, if the decoder 200 checks with code 218, then the coefficient 16 within each coefficient group 18 The bits of the corresponding encoded bit plane are read from the data stream 102 at reference numeral 220.
It follows a predetermined mapping rule for the decoder 200 to insert the bits from the bitstream 102 into the bitplane, so that the bits of the data stream 102, ie 122, are directly between the bitplanes indicated by GCLI and GTLI, ie. Read or decrypt.

In FIG. 9, it is further shown that the encoder 100 can optionally, within the data stream 102, notify the fact that the information 104 is related to this kind of significance indication, ie, for example, CSF rather than RSF. There is.
This instruction is shown as being optionally inserted into the data stream 102 by the encoder 100 at 250.

FIG. 11 shows an encoder 300 configured to use RSF instead of CSF.
The encoder 300 of FIG. 11 operates as follows.
In particular, the following description focuses on the difference from the operation of the encoder 100 of FIG.

The encoder 300 of FIG. 11 determines the GCLI predictors of all coefficient groups 18 of its groupset 40 at reference numeral 314, and whether all predictions are exactly matched by reference numeral 316, i.e., all within groupset 40. It works with the current groupset 40 by determining if the predicted residuals for the coefficient group 18 of are all zero. If so, the encoder 300 notifies this by setting RSF = 1 with reference numeral 318 in the significance information 104 of the data stream 102.
Here, instruction 250 indicates that instead of CSF notification, RSF notification is delivered in the information field 104 of the data stream 102, as indicated by instruction 250 in FIG.
However, if all GCLI predicted residuals are not 0, the RSF flag of this groupset 40 is set to 0 with sign 220, and the predicted residual of the GCLI value of the coefficient group 18 of the current groupset 40 is the sign. At 322, it is inserted into the data stream 102, i.e. into the field 118.
Regardless of whether RSF is set, it is checked if there is an untruncated coded bitplane for each coefficient group, and if so, it is inserted into the data stream 102 with code 326.

Encoders according to further embodiments, eg, according to both modes, i.e., according to FIG. 9 or FIG. 11, to determine which option, RSF or CSF, is preferred according to the sense of coding efficiency. Note that you can choose from to work.

FIG. 12 shows a decoder 400 capable of handling a data stream 102 containing an instruction 250, regardless of whether the instruction 250 indicates the use of CSF or RSF coding.
Although the reference numerals of FIG. 10 are reused, the information 104 is shown as "R / CSF" instead of "CSF" to show that the meaning of the flag of the information 104 depends on the indication 250.
The non-significant process 210 is performed by the decoder 400 only if the flag corresponding to the current groupset 40 is set and the CSF mode is simultaneously indicated by the instruction 250.
If not, a further difference in operating mode compared to FIG. 10 is predicted if check 402 is set for the R / CSF flag for the current groupset 40 and indication 250 indicates RSF mode. The fact that modification 216 is skipped by the decoder 400. If not, predictive modification 216 is performed.

It should be noted that with respect to FIG. 12, the decoder 400 of FIG. 12 is not very different from the decoder of FIG.
The ability to handle both RSF coding and CSF coding has little operational overhead.
On the other hand, for anyone who wants to install an encoder to generate a data stream 102 for feeding to the decoder 12, from the RSF option of FIG. 11 or the CSF option of FIG. 9, if the indication 250 is used. Opportunity to choose is provided. In this respect, the CSF option may have advantages with respect to the parallel processing function, but the RSF option in FIG. 11 is advantageous when the encoder is implemented in a sequential operation style, for example, in the form of FPGA. Please note.
In particular, the RSF setting depends on the prediction criteria base of the prediction in step 322, but the CSF setting can be done independently of other conversion factors unless one needs to know about GTLI, ie quantization.

With respect to FIGS. 9-12, the data stream 102 may be provided by the encoder with information or a flag as to whether the significant mode is used, and the information 104 and optionally the notification 250 are optionally data. Note that not present in stream 102 and instead all group sets 40 may be processed in normal mode.

<Definition and abbreviation>
These are some definitions used in the documentation.
GCLI: Maximum coded line index GCLI coefficient group: A group of wavelet coefficients represented by one GCLI value Escape GCLI: Not used for normal coding, but used to notify the decoder of specific conditions. GCLI value that can be significant GCLI: GCLI whose value is greater than zero
Non-significant GCLI: GCLI with a value of zero
GTLI: Maximum truncated line index Truncated GCLI: result of max (GCLI-GTLI, 0) Insignificant truncated GCLI: GCLI whose value is less than or equal to the coefficient group GTLI
GCLI Residual: The result of the prediction applied to the GCLI value. This requires a reference GCLI and a corresponding GTLI value. There are two variations, horizontal and vertical predictions.
Presynced: A group of coefficients of various subbands that contribute to a particular spatial region of the input image.
Scenario: Pre-synced-based defined quantization parameters that can be used to derive GTLI values for various wavelet subbands.
RSF: Residual significant flag, also known as insignificant flag [1].
SIG Group: A group of GCLI coefficient groups to which the SIG flag is assigned. Also known as a significant group.
CSF: Coefficient significance flag

[References]
[1] EP17162866.2, Decoder for decoded image data from data stream, encoder for encoded image data into a data data stream, data data stream
[2] intoPIX, "intoPIX Codec Submission for JPEG-XS CfP, Design Design", wg1m73019
[3] AMBROISE RENAUD; BUYSSCHERT CHARLES; PELLEGRIN PASCAL; ROUVROY GAEL, "Method and Device for display compression", EP2773122 A1
[4] AMBROISE RENAUD; BUYSSCHEERT CHARLES; PELLEGRIN PASCAL; ROUVROY GAEL, "Method and Device for Display Compression", US9332258 BB
[5] Jean-Baptiste Loent, "TICO Lightweight Codec Used in IP Networked or in SDI Infrastructure", SMPTE RDD 35: 2016
[6] Toshiaki Kojima, "LLVC-Low Latency Video Codec for Network Transfer", SMPTE RDD 34: 2015
[7] J. Kim and C. M. Kyung, "A Lossless Embedded Compression Usage Signing Bit Truncation for HD Video Coding", IEEE Transitions on Circuits and Systems and Systems and Systems

FIG. 13 shows an example of the pseudo code of the data stream 102.
In this pseudo code, the instruction 250 is transmitted within a parameter called "Rm".
Rm = 1 indicates the use of CSF coding mode, and the modification 502 of the GCLI predictor calculated at code 504 is sufficient in any case to ensure that the quantization threshold T tested at code 506 is not exceeded. By comprehensively setting the predicted residual Δm with reference numeral 500 so that the value becomes small, skipping of bit derivation is promoted.
Skipping the GCLI residual read from the data stream is done by rendering the predicted residual, i.e. the read of Δm, according to the predicted flag Z, based on the significance flag information of reference numeral 508.
Whether Rm is 0 or 1 does not affect this dependence of the predicted residual reading at code 510 on the significance flag at code 508.
If Rm is 0, that is, if the RSF mode is active, the predicted residual Δm is set to 0 with reference numeral 512.
Bit derivation of the coded bit plane is not shown in FIG. 13, but is done only for conversion coefficient groups where M is greater than 0.

Although some aspects have been described in the context of the device, these aspects also represent a description of the corresponding method, and it is clear that the block or device corresponds to a method step or feature of the method step.
Similarly, aspects described in the context of method steps also represent a description of the function of the corresponding block or item or corresponding device.
Some or all of the method steps may be performed by (or using) a hardware device such as a microprocessor, programmable computer, or electronic circuit.
In some embodiments, one or more of the most important method steps can be performed by such a device.

The coded data stream of the present invention can be stored in a digital storage medium, or can be transmitted by a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

Depending on the specific implementation requirements, embodiments of the present invention can be implemented in hardware or software.
This embodiment uses a digital storage medium, such as a floppy disk, DVD, Blu-ray, CD, ROM, PROM, EPROM, EEPROM, or flash memory, which stores electronically readable control signals. These can work with (or can work with) a programmable computer system so that each method is performed.
Therefore, the digital storage medium may be computer readable.

Some embodiments according to the invention are data carriers having electronically readable control signals capable of cooperating with a programmable computer system such that one of the methods described herein is performed. including.

In general, embodiments of the present invention can be implemented as a computer program product with program code, which operates to perform one of the methods when the computer program product is run on a computer. ..
The program code may be stored, for example, in a machine-readable carrier.

Other embodiments include a computer program stored in a machine-readable carrier for performing one of the methods described herein.

In other words, therefore, an embodiment of the method of the invention is a computer program having program code for performing one of the methods described herein when the computer program is executed on a computer.

Accordingly, further embodiments of the methods of the invention include a computer program for performing one of the methods described herein, the data carrier (or digital storage medium, or computer-readable medium) on which it is recorded. Is.
Data carriers, digital storage media, or recorded media are usually tangible and / or non-temporary.

Therefore, a further embodiment of the method of the invention is a sequence of data streams or signals representing a computer program for performing one of the methods described herein.
A data stream or sequence of signals may be configured to be transferred over a data communication connection, such as the Internet.

Further embodiments include processing means configured or adapted to perform one of the methods described herein, such as a computer, or a programmable logic device.

Further embodiments include a computer on which a computer program for performing one of the methods described herein is installed.

A further embodiment according to the invention is a device or system configured to transfer (eg, electronically or optically) a computer program to a receiver to perform one of the methods described herein. Including.
The receiver may be, for example, a computer, a mobile device, a memory device, or the like.
The device or system may include, for example, a file server for transferring computer programs to the receiver.

In some embodiments, programmable logic devices (eg, field programmable gate arrays) may be used to perform some or all of the functions of the methods described herein.
In some embodiments, the field programmable gate array may work with a microprocessor to perform one of the methods described herein.
In general, the method is preferably performed by any hardware device.

The devices described herein can be implemented using hardware devices, using computers, or using a combination of hardware devices and computers.

The devices described herein, or any component of the devices described herein, may be implemented, at least in part, in hardware and / or software.

The methods described herein may be performed using hardware equipment, using a computer, or using a combination of hardware equipment and computer.

The methods described herein, or any component of the device described herein, may be performed at least partially by hardware and / or software.

The above embodiments are merely examples of the principles of the present invention.
It should be understood that modifications and changes to the arrangements and details described herein will be apparent to those skilled in the art.
Therefore, it is intended to be limited only by the imminent claims, not by the specific details presented as the description and description of the embodiments herein.

は［２］で使用される値と同じ値であるため、複雑さについてはこれ以上説明しない。

の計算は、コンパレータ（＜＝５ＬＵＴ）およびサブバンドごとに４ビットの１つのレジスタによって可能である。
さらに、ＧＣＬＩ値を１つの有意グループだけ遅延させることにより、初期予算計算が簡素化される。
１つの垂直ウェーブレット分解レベル（３×８サブバンド）の場合、これには３×８×８×４＝７６８ビットが必要である。
ザイリンクスの場合、これは２×４８＝９６ＬＵＴ、またはアルテラデバイスの２ ＭＬＡＢブロックに対応する。

Is the same value used in [2], so complexity will not be discussed further.

The calculation of is possible with a comparator (<= 5 LUT) and one register with 4 bits per subband.
In addition, delaying the GCLI value by one significant group simplifies initial budgeting.
For one vertical wavelet decomposition level (3x8 subband), this requires 3x8x8x4 = 768 bits.
For Xilinx, this corresponds to 2 x 48 = 96 LUT, or 2 MLAB blocks for Altera devices.

（１）
現在のＧＴＬＩと基準ＧＴＬＩのｔ_ｉおよびｔ_ｉ−１の両方が等しい場合、式（１）は次のように単純化される。

（２）
予算の節約は

についてのみ発生することがわかっているため、式（１）から次式を得る。
δ_ｉ＝ｔ_ｉ−ｍａｘ（ｇ_ｉ−１，ｍａｘ（ｔ_ｉ−１，ｔ_ｉ））
したがって、予算の節約は、ｇ_ｉ−１と

、さらにパラメータｔ_ｉ−１とｔ_ｉに一意に依存するため、簡単に計算できる。 The following equation is used in this prediction method.

(1)
If both _{t i} and _{t i-1} of the current GTLI and reference GTLI are equal, equation (1) is simplified as follows.

(2)
Budget savings

Since it is known that this occurs only in the above equation (1), the following equation is obtained.
_{δ i = t i -max (g} i-1, max (t i-1, t i))
Therefore, the budget savings are g _i-1

, To uniquely dependent on further parameters _{t i-1} and _{t i,} can be easily calculated.

Claims (45)

A decoder configured to decode the conversion factor (16) from the data stream (102).
The conversion coefficients are grouped into coefficient groups (18).
The coefficient groups (18) are grouped into group sets (40).
The decoder
A significant coding mode indication (250) is derived from the data stream (102),
From the data stream (102), a first subset of the groupset (40) in which the significant coding mode is not used and a second subset of the groupset (40) in which the significant coding mode is used are identified. Information (104) to be derived
For each groupset of the first subset, each coefficient group of each of the groupsets
Derivation of the first prediction of the first set of coded bit planes based on the first previously decoded coefficient group (214) and
Modifying the first prediction using the first prediction residual (118) derived from the data stream to obtain the modified prediction of the first set of coded bit planes. (216) and
Deriving the bits of each of the respective coefficient groups in the modified prediction of the first set of coded bitplanes from the data stream (220) and thereby the first set of coded bitplanes. Identify and
For each groupset of the second subset and each coefficient group of the respective groupset, if the significant coding mode is the first mode,
Identifying the second set of coded bitplanes by deriving a second prediction of a second set of coded bitplanes based on a second previously decoded coefficient group (214). ,
Derived from the data stream the bits of each of the coefficient groups in the prediction of the second set of encoded bit planes (220).
For each groupset of the second subset, if the significant coding mode is the second mode,
A decoder configured to inherit that for each coefficient group in each group set, the coefficients of the respective coefficient groups are insignificant (210). The decoder according to claim 1, wherein the conversion coefficient forms one lower part of a plurality of lower parts of the spectral decomposition of the image (12). The decoder according to claim 2, wherein the decoder is configured to derive the significant coding mode from the data stream for each lower portion. The decoder according to claim 3, wherein the lower portion is a subband or group of conversion coefficients with respect to the corresponding spatial region (32) of the spatial region in which the image (12) is subdivided. It is configured to derive truncation information (108) from the data stream that indicates one or more sets of truncated least significant bit planes in each coefficient group.
The first prediction, the modified prediction, and the second prediction determine the most significant bitplane index for the most significant bitplane of the set of one or more truncated least significant bitplanes. The decoder according to any one of claims 1 to 4, as shown. Coefficients of conversion factors for the corresponding spatial region (32) of the spatial region in which the image (12) is subdivided. Group Row, subband, or group granularity of the one or more truncated least significant bit planes. The decoder according to claim 5, configured to derive the truncated information (108) indicating a set from the data stream. The decoder according to any one of claims 1 to 6, wherein the conversion coefficient is a spectral coefficient of the spectral decomposition (10) of the image (12). The decoder according to any one of claims 1 to 7, wherein the conversion coefficient is a DCT or a wavelet coefficient. The conversion coefficient is the spectral coefficient of the spectral decomposition (10) of the image (12) into subbands, and the conversion coefficient (16) in one coefficient group (18) belongs to the same subband. The decoder according to any one of claims 1 to 8, which is grouped into a coefficient group (18). It is configured to derive the information (104) from the data stream as one flag for each groupset (40).
The groupset (40) in which the flag assumes the first state belongs to the first subset of the groupset.
The decoder according to any one of claims 1 to 9, wherein the groupset in which the flag assumes a second state belongs to the second subset of the groupset. The decoder according to any one of claims 1 to 10, configured to perform the derivation (220) of the bit at a coding rate of 1. The decoder according to any one of claims 1 to 11, configured to set a non-significant coefficient to zero or pseudo-noise. An encoder configured to encode the conversion factor (16) into a data stream (102).
The conversion coefficients are grouped into coefficient groups (18).
The coefficient groups are grouped into a group set (40).
The encoder
The first mode or the second mode, the significant coding mode (250), is notified in the data stream (102).
Information (104) that identifies the first subset of the groupset in which the significant coding mode is not used and the second subset of the groupset in which the significant coding mode is used is provided in the data stream (102). Insert and
For each groupset of the first subset, each coefficient group of each of the groupsets
Derivation of the first prediction of said first set of encoded bit planes based on the first previously encoded coefficient group (114), and
In order to obtain the modified predictions of the first set of coded bit planes, a first prediction residual (118) for modifying the first prediction is inserted into the data stream (102). That (116) and
By inserting into the data stream the bits of each of the coefficient groups in the modified prediction of the first set of coded bit planes (120), of the coded bit plane in the data stream. Identify the first set and
For each coefficient group of the respective groupset of each groupset of the second subset, if the significant coding mode is the first mode.
Identifying the second set of coded bitplanes by deriving a second prediction of a second set of coded bitplanes based on a second previously encoded coefficient group (114). And
It is configured to insert (120) the bits in the prediction of the second set of coded bit planes into the data stream.
The significant coding mode, which is the second mode, notifies that the coefficient of each coefficient group is not significant for each groupset of the second subset and each coefficient group of each groupset. .. The encoder according to claim 13, wherein the conversion coefficient forms one lower part of a plurality of lower parts of the spectral decomposition of the image. 14. The encoder according to claim 14, wherein the encoder is configured to select and change the significant coding mode from at least the first and second modes for each lower portion by notification in the data stream. Encoder. The encoder according to claim 15, wherein the lower portion is a subband or group of conversion coefficients for the corresponding spatial region of the spatial region in which the image is subdivided. It is configured to insert truncation information into the data stream that indicates one or more sets of truncated least significant bit planes in each coefficient group.
The first prediction, the modified prediction, and the second prediction determine the most significant bitplane index for the most significant bitplane of the set of one or more truncated least significant bitplanes. The encoder according to any one of claims 13 to 16, as shown. Coefficients of conversion factors for the corresponding spatial regions of the spatial region in which the image is subdivided. The truncated information indicating the set of the one or more truncated least significant bit planes, with the granularity of the group row, subband, or group. 17. The encoder according to claim 17, wherein the encoder is configured to be inserted into the data stream. The encoder is configured to select the significant coding mode from at least the first and second modes by default, or
The encoder tests at least the first set of coding modes, including the first and second modes, and selects the best coding mode from the set of coding modes according to a predetermined criterion. It is configured to select the significant coding mode from the first and second modes.
The encoder according to any one of claims 13 to 18. 19. The encoder according to claim 19, wherein the predetermined reference depends on the coding rate and / or the coding distortion. The encoder according to any one of claims 13 to 20, wherein the conversion coefficient is a spectral coefficient for spectral decomposition of an image. The encoder according to any one of claims 13 to 21, wherein the conversion coefficient is a DCT or a wavelet coefficient. The conversion coefficient is a spectral coefficient of spectral decomposition into subbands of an image, and is grouped into the coefficient group in such a manner that the conversion coefficient in one coefficient group belongs to the same subband, according to claim 13. The encoder according to any one of 22. It is configured to insert the information into the data stream as one flag for each groupset.
A groupset in which the flag assumes a first state belongs to the first subset of the groupset.
The encoder according to any one of claims 13 to 23, wherein the groupset in which the flag assumes a second state belongs to the second subset of the groupset. The encoder according to any one of claims 13 to 24, which is configured to perform the insertion of the bit at a code rate of 1. The encoder according to any one of claims 13 to 25, wherein the non-significant coefficient is notified to be set to zero or pseudo-noise. A decoder configured to decode the conversion factor (16) from the data stream (102).
The conversion coefficients (16) are grouped into coefficient groups (18).
The coefficient groups are grouped into a group set (40).
The decoder
Information (104) from the data stream (102) that identifies a first subset of the groupset in which the significant coding mode is not used and a second subset of the groupset in which the significant coding mode is used. Derived and
For each groupset of the first subset, each coefficient group of each of the groupsets
Derivation of the first prediction of the first set of coded bit planes based on the first previously decoded coefficient group (214) and
Modifying the first prediction using the first prediction residual (118) derived from the data stream to obtain the modified prediction of the first set of coded bit planes. (216) and
By deriving from the data stream the bits of each of the respective coefficient groups in the modified prediction of the first set of coded bitplanes at a code rate of 1 (220), of the coded bitplane. Identify the first set and
A decoder configured to inherit (210) that for each groupset of the second subset, for each coefficient group of the respective groupset, the coefficients of the respective coefficient groups are not significant. Whether or not the significant coding mode is applied is derived from the data stream (102).
27. The 27th aspect of claim 27, wherein if the significant coding mode is not applied, the derivation of the information from the data stream is skipped and the second subset is presumed to be empty. decoder. The conversion factor forms one subpart of the plurality of subparts of the spectral decomposition of the image.
28. The decoder according to claim 28, wherein the decoder is configured to derive from the data stream whether or not the significant coding mode is applied for each lower portion. 29. The decoder according to claim 29, wherein the lower portion is a subband or group of conversion coefficients for the corresponding spatial region of the spatial region in which the image is subdivided. It is configured to derive truncation information from the data stream that indicates one or more sets of truncated least significant bit planes in each coefficient group.
From claim 27, the first prediction and the modified prediction indicate the most significant bitplane index for the most significant bitplane in the set of one or more truncated least significant bitplanes. The decoder according to any one of 30. Coefficients of conversion factors for the corresponding spatial regions of the spatial region in which the image is subdivided. The truncated information indicating the set of the one or more truncated least significant bit planes, with the granularity of the group rows, subbands, or groups. 31. The decoder according to claim 31, which is configured to derive the above data stream. The decoder according to any one of claims 27 to 32, wherein the conversion coefficient is a spectral coefficient for spectral decomposition of an image. The decoder according to any one of claims 27 to 33, wherein the transform coefficient is a DCT or a wavelet coefficient. The conversion coefficient is a spectral coefficient of spectral decomposition into subbands of an image.
The decoder according to any one of claims 27 to 34, wherein the conversion coefficients in one coefficient group are grouped into the coefficient groups in such a manner that they belong to the same subband. It is configured to derive the information from the data stream as one flag for each groupset.
A groupset in which the flag assumes a first state belongs to the first subset of the groupset.
The decoder according to any one of claims 27 to 35, wherein the groupset in which the flag assumes a second state belongs to the first subset of the groupset. The decoder according to any one of claims 27 to 36, configured to perform the derivation of the first predicted residual using a unidirectional VLC code. The decoder according to any one of claims 27 to 37, configured to set a non-significant coefficient to zero or pseudo-noise. An encoder configured to encode the conversion factor into a data stream.
The conversion coefficients are grouped into coefficient groups and
The coefficient groups are grouped into group sets and
The encoder
Information that identifies a first subset of the groupset in which the significant coding mode is not used and a second subset of the groupset in which the significant coding mode is used is inserted into the data stream.
For each groupset of the first subset, each coefficient group of each of the groupsets
Identifying the first set of coded bitplanes by deriving a first prediction of a first set of coded bitplanes based on a first previously decoded coefficient group (214). ,
In order to obtain the modified predictions of the first set of coded bit planes, a first prediction residual for modifying the first prediction is inserted into the data stream.
Bits of each of the coefficient groups in the modified prediction of the first set of coded bit planes are configured to be inserted into the data stream at a code rate of 1.
For each groupset of the second subset, each coefficient group of each groupset, the coefficients of each of the coefficient groups are not significant, the encoder. A method for decoding the conversion factor (16) from the data stream (102).
The conversion coefficients are grouped into coefficient groups (18).
The coefficient groups (18) are grouped into group sets (40).
The method is
A step of deriving a significant coding mode instruction (250) from the data stream (102), and
From the data stream (102), a first subset of the groupset (40) in which the significant coding mode is not used and a second subset of the groupset (40) in which the significant coding mode is used are identified. Steps to derive the information (104) to be
For each group set of the first subset, each coefficient group of each of the group sets, a first prediction of a first set of encoded bitplanes based on a first previously decoded coefficient group. Derivation (214) and using the first prediction residual (118) derived from the data stream to obtain the modified predictions of the first set of encoded bitplanes. Modifying the first prediction (216) and deriving the bits of each of the respective coefficient groups within the modified prediction of the first set of encoded bitplanes from the data stream (220). To identify the first set of encoded bitplanes by
For each group set of the second subset, each coefficient group of the respective group set, if the significant coding mode is the first mode, based on the second previously decoded coefficient group. By deriving a second prediction of a second set of coded bit planes (214), the step of identifying the second set of coded bit planes.
A step (220) of deriving the bits of each of the coefficient groups in the prediction of the second set of coded bit planes from the data stream.
For each groupset of the second subset, if the significant coding mode is the second mode, for each coefficient group of each groupset, inherit that the coefficients of each coefficient group are not significant. Step (210) to do
Including methods. A method for encoding the conversion factor (16) into a data stream (102).
The conversion coefficients are grouped into coefficient groups (18).
The coefficient groups are grouped into a group set (40).
The method is
A step of notifying the significant coding mode (250) of the first mode or the second mode in the data stream (102), and
Information (104) that identifies the first subset of the groupset in which the significant coding mode is not used and the second subset of the groupset in which the significant coding mode is used is provided in the data stream (102). Steps to insert and
For each group set of the first subset, each coefficient group of each of the group sets, a first prediction of a first set of encoded bitplanes based on a first previously encoded coefficient group. (114) and a first prediction residual (118) for modifying the first prediction in order to obtain the modified prediction of the first set of encoded bit planes. Inserting into the data stream (102) (116) and inserting the bits of each of the respective coefficient groups in the modified prediction of the first set of encoded bit planes into the data stream (120). ), And the step of identifying the first set of encoded bit planes in the data stream.
For each group set of the second subset, each coefficient group of each of the group sets, if the significant coding mode is the first mode, then the second previously encoded coefficient group Based on this, deriving a second prediction of a second set of coded bit planes (114) and inserting the bits in the prediction of the second set of coded bit planes into the data stream. (120), including the step of identifying the second set of encoded bit planes.
The significant coding mode, which is the second mode, is a method of notifying that the coefficient of each coefficient group is not significant for each groupset of the second subset and each coefficient group of each groupset. .. A method for decoding the conversion factor (16) from the data stream (102).
The conversion coefficients (16) are grouped into coefficient groups (18).
The coefficient groups are grouped into a group set (40).
The method is
Information (104) from the data stream (102) that identifies a first subset of the groupset in which the significant coding mode is not used and a second subset of the groupset in which the significant coding mode is used. Steps to derive and
For each group set of the first subset, each coefficient group of each of the group sets, the first prediction of the first set of coded bitplanes is made based on the first previously decoded coefficient group. Derivation (214) and using the first prediction residual (118) derived from the data stream to obtain the modified predictions of the first set of coded bit planes. Modifying the first prediction (216) and deriving from the data stream at a coding factor 1 the bits of each of the respective coefficient groups in the modified prediction of the first set of coded bit planes. By that (220), the step of identifying the first set of encoded bit planes,
A method comprising, for each groupset of the second subset, for each coefficient group of each groupset, a step (210) of inheriting that the coefficients of each of the coefficient groups are not significant. A method for encoding the conversion factor into a data stream,
The conversion coefficients are grouped into coefficient groups and
The coefficient groups are grouped into group sets and
The method is
A step of inserting into the data stream information that identifies a first subset of the groupset in which the significance coding mode is not used and a second subset of the groupset in which the significance coding mode is used.
For each group set of the first subset, each coefficient group of each of the group sets, the first prediction of the first set of coded bitplanes is made based on the first previously decoded coefficient group. Insert a first prediction residual into the data stream to modify the first prediction in order to derive and obtain the modified prediction of the first set of encoded bit planes. By inserting the bits of each of the coefficient groups in the modified prediction of the first set of coded bit planes into the data stream at a coding rate of 1, the coded bit plane. Including,
A method, wherein for each groupset of the second subset, for each coefficient group of each groupset, the coefficients of each of the coefficient groups are not significant. A data stream generated by the method of claim 41 or claim 43. A computer program having a program code for executing the method according to any one of claims 40 to 43 when the computer is executed.

Images